Summary:
The Android ImageEditingManager is inefficient and slow when cropping images. It loads the full resolution image into memory and then crops it. This leads to slow performance and occasional OutOfMemory Exceptions.
[BitmapRegionDecoder](https://developer.android.com/reference/android/graphics/BitmapRegionDecoder.html) can be used to crop without needing to load the full resolution image into memory. Using it is much more efficient and much faster.
Relevant issue: https://github.com/facebook/react-native/issues/10470
Attempt to crop a very large image (2000x2000) on Android. With this change, the crop should happen almost instantly. On the master branch, it should take 2-3 seconds (and might run out of memory).
Please let me know if there's anything else I can provide.
Closes https://github.com/facebook/react-native/pull/15439
Differential Revision: D5628223
Pulled By: shergin
fbshipit-source-id: bf314e76134cd015380968ec4533225e724c4b26
Summary:
Instead of preventing events from working when not on the UI Thread we can just dispatch to it instead.
**Test plan**
Tested manually that animated events still work in RNTester
Closes https://github.com/facebook/react-native/pull/15953
Differential Revision: D5909816
Pulled By: shergin
fbshipit-source-id: 48d02b6aa9f2bc3bcb638e8852fccaac3f205276
Summary:
This problem tries to solve issue #14244
Implemented a fetch request to a TLSv1 server. It now successfully resolves TLS handshake.
Tested same fetch to TLSv1.2 server and still successfully resolves TLS handshake.
Closes https://github.com/facebook/react-native/pull/14245
Differential Revision: D5898689
Pulled By: shergin
fbshipit-source-id: 8766ebe6909443367651ab868aa5ff62747cd906
Summary:
Following the migration guide. Let's see what happens here.
Closes https://github.com/facebook/react-native/pull/14955
Differential Revision: D5877682
Pulled By: hramos
fbshipit-source-id: 2a40560120b5d8d28bc6c52cc5e5916fd1bba336
Summary:
As I was working on mimicking iOS animations for my ongoing work with `react-navigation`, one task I had was to match the "push from right" animation that is common in UINavigationController.
I was able to grab the exact animation values for this animation with some LLDB magic, and found that the screen is animated using a `CASpringAnimation` with the parameters:
- stiffness: 1000
- damping: 500
- mass: 3
After spending a considerable amount of time attempting to replicate the spring created with these values by CASpringAnimation by specifying values for tension and friction in the current `Animated.spring` implementation, I was unable to come up with mathematically equivalent values that could replicate the spring _exactly_.
After doing some research, I ended up disassembling the QuartzCore framework, reading the assembly, and determined that Apple's implementation of `CASpringAnimation` does not use an integrated, numerical animation model as we do in Animated.spring, but instead solved for the closed form of the equations that govern damped harmonic oscillation (the differential equations themselves are [here](https://en.wikipedia.org/wiki/Harmonic_oscillator#Damped_harmonic_oscillator), and a paper describing the math to arrive at the closed-form solution to the second-order ODE that describes the DHO is [here](http://planetmath.org/sites/default/files/texpdf/39745.pdf)).
Though we can get the currently implemented RK4 integration close by tweaking some values, it is, the current model is at it's core, an approximation. It seemed that if I wanted to implement the `CASpringAnimation` behavior _exactly_, I needed to implement the analytical model (as is implemented in `CASpringAnimation`) in `Animated`.
We add three new optional parameters to `Animated.spring` (to both the JS and native implementations):
- `stiffness`, a value describing the spring's stiffness coefficient
- `damping`, a value defining how the spring's motion should be damped due to the forces of friction (technically called the _viscous damping coefficient_).
- `mass`, a value describing the mass of the object attached to the end of the simulated spring
Just like if a developer were to specify `bounciness`/`speed` and `tension`/`friction` in the same config, specifying any of these new parameters while also specifying the aforementioned config values will cause an error to be thrown.
~Defaults for `Animated.spring` across all three implementations (JS/iOS/Android) stay the same, so this is intended to be *a non-breaking change*.~
~If `stiffness`, `damping`, or `mass` are provided in the config, we switch to animating the spring with the new damped harmonic oscillator model (`DHO` as described in the code).~
We replace the old RK4 integration implementation with our new analytic implementation. Tension/friction nicely correspond directly to stiffness/damping with the mass of the spring locked at 1. This is intended to be *a non-breaking change*, but there may be very slight differences in people's springs (maybe not even noticeable to the naked eye), given the fact that this implementation is more accurate.
The DHO animation algorithm will calculate the _position_ of the spring at time _t_ explicitly and in an analytical fashion, and use this calculation to update the animation's value. It will also analytically calculate the velocity at time _t_, so as to allow animated value tracking to continue to work as expected.
Also, docs have been updated to cover the new configuration options (and also I added docs for Animated configuration options that were missing, such as `restDisplacementThreshold`, etc).
Run tests. Run "Animated Gratuitous App" and "NativeAnimation" example in RNTester.
Closes https://github.com/facebook/react-native/pull/15322
Differential Revision: D5794791
Pulled By: hramos
fbshipit-source-id: 58ed9e134a097e321c85c417a142576f6a8952f8
Summary:
Fixes https://github.com/facebook/react-native/issues/15863 on master. Behavior of `onSubmitEditing` is now consistent between iOS and Android. Tapping the submit button in a TextInput dispatches the event precisely when doing so does not make a newline (when blurOnSubmit is true or multiline is false).
1. Run this app on iOS and Android:
```
// flow
import React, { Component } from 'react';
import {
StyleSheet,
TextInput,
View
} from 'react-native';
type State = {
toggled: boolean
};
type Props = {
blurOnSubmit: boolean,
multiline: boolean
};
class ToggleColorInput extends Component<Props, State> {
state: State = {
toggled: false
};
props: Props;
toggle = () => {
this.setState({
toggled: !this.state.toggled
});
}
render() {
return (
<TextInput
blurOnSubmit={this.props.blurOnSubmit}
multiline={this.props.multiline}
onSubmitEditing={this.toggle}
style={[styles.textInput, {backgroundColor: this.state.toggled ? 'blue' : 'azure'}]}
underlineColorAndroid='transparent'
/>
)
}
}
export default class App extends Component<{}> {
render() {
return (
<View>
<ToggleColorInput blurOnSubmit={true} multiline={true} />
<ToggleColorInput blurOnSubmit={true} multiline={false} />
<ToggleColorInput blurOnSubmit={false} multiline={true} />
<ToggleColorInput blurOnSubmit={false} multiline={false} />
</View>
);
}
}
const styles = StyleSheet.create({
textInput: {
height: 75,
borderWidth: 1,
borderColor: 'black'
}
});
```
2. You see four TextInputs, with each combination of the `blurOnSubmit` and `multiline` properties. For each TextInput, type some text and tap the submit button.
3. The TextInputs in this test will toggle background color when they emit an `onSubmitEditing` event. Verify the following behavior on each platform:
* blurOnSubmit && isMultiline => Submit event emitted, blurred, no newline inserted
* blurOnSubmit && !isMultiline => Submit event emitted, blurred
* !blurOnSubmit && isMultiline => Submit event emitted, newline inserted
* !blurOnSubmit && !isMultiline => Submit event emitted
Closes https://github.com/facebook/react-native/pull/16040
Differential Revision: D5877401
Pulled By: shergin
fbshipit-source-id: 741bcc06d8b69d7025f2cb42dd0bee4fa01cd88e
Summary:
<!--
Thank you for sending the PR! We appreciate you spending the time to work on these changes.
Help us understand your motivation by explaining why you decided to make this change.
You can learn more about contributing to React Native here: http://facebook.github.io/react-native/docs/contributing.html
Happy contributing!
-->
Headless tasks are required to run in the main thread, however due to the nature of the React context creation flow, the handler may be returned outside of the main thread, causing the HeadlessJsTaskContext to throw an exception.
Swipe out the app. send push notification from a server that starts a HeadlessJsTaskService
Closes https://github.com/facebook/react-native/pull/15940
Differential Revision: D5852448
Pulled By: foghina
fbshipit-source-id: 54c58a1eb7434dd5de5c39c28f6e068ed15ead9d
Summary:
When we convert nested <Text> components to Spannable object we must enforce the order of spans somehow,
otherwise we will have Spannable object with unpredictable order of spans, which will produce unpredictalbe text layout.
We can do it only using `Spannable.SPAN_PRIORITY` feature because Spannable objects do not maintain the order of spans internally.
We also have to fix this to implement autoexpandable <TextInput>.
Reviewed By: achen1
Differential Revision: D5811172
fbshipit-source-id: 5bc68b869e58aba27d6986581af9fe3343d116a7
Summary:
**Problem:**
It was observed that in [this code path](https://github.com/facebook/react-native/blob/master/ReactAndroid/src/main/java/com/facebook/react/views/scroll/ReactHorizontalScrollView.java#L292) (i.e. horizontal, paging-enabled scroll view) if you tried to programmatically call the scrollTo method within ~1s of the onMomentumScrollEnd event (which should only be called after all scrolling has ended), the scrollView would scroll to the new location, and then scroll BACK to the original location.
For example, assume you had released the scrollView at location B, and the nearest page boundary is A. Then, 1000ms later, you call scrollTo position C. The order of operations would be:
1. Begin scrolling to A from position B (as it is the nearest page boundary)
2. Reach position A
3. Fire onMomentumScrollEnd
4. 1000ms later call scrollTo C
5. scrollView scrolls to C
6. scrollView scrolls BACK to position A (for no apparent reason).
**Reason:**
I suspect this is because the [smoothlyScrollTo](https://github.com/facebook/react-native/blob/master/ReactAndroid/src/main/java/com/facebook/react/views/scroll/ReactHorizontalScrollView.java#L292) will continue to animate towards A, but the [scrollEvents will not fire](f954f3d9b6/ReactAndroid/src/main/java/com/facebook/react/views/scroll/OnScrollDispatchHelper.java (L45)) as they are too close to each other. So the true order of events is:
1. Begin scrolling to A from position B (as it is the nearest page boundary)
[begin smoothlyScrollTo]
[scroll towards position A]
[mActivelyScrolling is true]
2. Reach position A
[mActivelyScrolling is true]
[scroll towards position A]
[mActivelyScrolling is false, as there is another scrollEvent, but because it is close enough to the same location it is ignored]
3. Fire onMomentumScrollEnd
4. 1000ms later call scrollTo C
[scroll towards position C]
5. scrollView scrolls to C
[scroll towards position A as the original smoothlyScrollTo animation was never completed]
6. scrollView scrolls BACK to position A.
This is an untested hypothesis, but seems to explain the behavior, and the solution is more semantically correct anyway. If there is an easy way to rebuild the android binaries happy to test it myself! Just let me know!
**Solution:**
Move the mActivelyAnimating outside the mOnScrollDispatchHelper.onScrollChanged helper, because the HorizontalScrollView event should be considered to be animating so long as onScrollChanged events are being fired.
Closes https://github.com/facebook/react-native/pull/15146
Reviewed By: AaaChiuuu
Differential Revision: D5792026
Pulled By: tomasreimers
fbshipit-source-id: 9654fda038d4a687cc32f4c32dc312baa34627ed
Summary:
Abstract class `ReactBaseTextShadowNode` was decoupled from `ReactTextShadowNode` to separate two goals/roles:
* `ReactBaseTextShadowNode` represents spanned `<Text>` nodes, which can bear text attributes (both `RCTText` and `RCTVirtualText`);
* `ReactTextShadowNode` represents anchor `<Text>` view in Yoga terms, which can bear layout attributes (`RCTText` and `RCTTextInput`).
`ReactVirtualTextShadowNode` now inherits `ReactBaseTextShadowNode`.
The same architectural changes was applited to view managers.
Why?
* This is just a better architecture which represents the nature of this objects.
* Bunch of "negative" logic which turn off excessive features for some suclasses was removed.
* Memory efficiency.
* Now we can improve `<TextInput>` component using right inheritance. Yay!
Reviewed By: achen1
Differential Revision: D5715830
fbshipit-source-id: ecc0764a03b5b7586fe77ad31f149cd840f4da41
Summary:
ReactRawTextShadowNode represents "raw" text node (aka textContent), so it cannot have layout or styling, it is just a line of text, a pure string.
So, we must not interit it from HUGE ReactTextShadowNode (which represents <Text> node with ReactRawTextShadowNode instance inside).
We need this change to make future fancy (and valuable from product perspective) refactoring possible. Stay tuned!
Reviewed By: achen1
Differential Revision: D5712961
fbshipit-source-id: 009e48046bdf34ddfd40b93175934969af64b98b
Summary:
We have to have a way to explicitly enforce the fact that some nodes cannot have Yoga child nodes.
Previously we relied on `isMeasureDefined()`, which is actually special case (so it does not cover all possible cases).
Reviewed By: AaaChiuuu
Differential Revision: D5647855
fbshipit-source-id: 59591be61ef62c61eb98748d44bb28b878f713fc
Summary:
This code related to velocity would cause some scroll events to be skipped and caused some jitter for sticky headers. Not sure if there is a better fix but removing this does fix missing events and the velocity calculation still seems to be working.
**Test plan**
Tested that sticky headers now work properly on Android, tested that velocity calculation still seem to work.
Closes https://github.com/facebook/react-native/pull/15761
Differential Revision: D5760820
Pulled By: sahrens
fbshipit-source-id: 562b5f606bdc5452ca3d85efb5e0e3e7db224891
Kevin Doherty
faifai21
Guizmo Martinez
Janic Duplessis
Paul van Staden
Andrew Gallagher
Alex Dvornikov
Charles Dick
Jia Li
Hector Ramos
Alexey Lang
Adam Miskiewicz
Adam King
Aaron Chiu
Andrew Chen (Eng)
Kathy Gray
Dmitry Zakharov
David Vacca
Kang Zhang
Tomas Reimers
Max Sherman
Michał Gregorczyk
Ori Harel
Valentin Shergin
Brian Vaughn
Ramanpreet Nara
Tomas Reimers
Spencer Ahrens
Andrew Jack
Hongtan Du
Michael Lee
Marc Horowitz
